AI Changed the Technology. It Didn’t Change the Rules.

After years of selling construction technology, Jeff Stieber thinks this may be the easiest time ever to get a buyer’s attention—and one of the hardest times to actually close the deal.

For most of my career in construction technology, we’ve been telling ourselves some version of the same story: construction is behind. It’s slow to adopt technology, resistant to change and stuck doing things the way they’ve always been done.

There was certainly some truth to that once, but, lately, I’ve been wondering if we’ve repeated it for so long that we haven’t noticed the story changing.

Jeff Stieber certainly thinks it has.

Jeff has spent a good chunk of his career selling technology into construction, including stops at Procore, Briq and Trunk Tools. He’s worn all of the hats. He’s been an individual contributor, managed strategic accounts and led sales teams. Today, through Stieber Advisory Group, he advises construction technology companies and recruits GTM talent. In other words, he’s watched several generations of construction technology come and go.

So, when I sat down with Jeff recently, I wanted to talk about what has actually changed, and, particularly in this moment of AI mania, what hasn’t.

Pretty quickly, he said something that I think perfectly captures where construction technology is right now:

“This is probably the easiest time in my career to get conversations for sales reps. I think it’s the hardest time to close deals because buyers are confused.”

We spent decades trying to get construction companies interested in technology. Now they may have too much of it to choose from. It’s a bit of a monster of our own creation.

Construction isn’t ignoring technology anymore

When Jeff first entered ConTech at Procore, VDC and BIM teams tended to be the people playing with the “shiny objects.” Over time, he watched many of those roles evolve into technology roles and eventually into innovation organizations.

That evolution matters. Technology decisions increasingly aren’t random experiments happening on individual projects. Larger construction organizations have people specifically charged with evaluating technology, developing strategies around it and figuring out how it fits across the business.

Now AI has accelerated the whole thing, and Jeff went so far as to tell me that he doesn’t think we should call construction a technology laggard anymore.

“Anybody that says construction is a laggard industry is wrong. Just look at everybody. Money’s flowing in and people are spending money.”

I mostly agree with him, although the data makes the distinction a little more nuanced.

RICS surveyed more than 2,200 construction professionals globally in 2025 and found that 45% reported no AI implementation, while another 34% were still in early pilot phases. Just under 12% reported regular use in specific processes. That certainly doesn’t sound like an industry moving at Silicon Valley speed.

But, look at why adoption is slow. The biggest reported barriers were lack of skilled personnel (46%), integration with existing systems (37%), data quality and availability (30%), implementation costs (29%) and unclear ROI (28%). Resistance to change came in at just 20%.

That distinction is important. Maybe construction can still be slow to implement technology without being resistant to technology itself. Those aren’t the same thing.

In fact, the same RICS research found substantial optimism about AI’s potential: 69% of project managers and 67% of quantity surveying and construction professionals agreed that AI will help surveyors deliver greater value in the future.

The appetite is there. The harder question is what actually deserves to be adopted.

AI has made buying technology more complicated

This is where Jeff thinks things have become particularly interesting.

For almost any problem a construction company wants to solve today, he says buyers are increasingly faced with three choices:

“Will my legacy application have a solution for it? My Procore, Autodesk—are they going to have a module to support it? Can I vibecode or build this myself? Or do I need to go get a new point solution that’s going to solve it?”

And, as Jeff points out, the answer could legitimately be any of the three.

That’s a very different problem from the one ConTech faced a decade ago. The old challenge was often convincing someone that software could solve the problem at all. The new challenge may be convincing them that your software is the one worth betting on.

AI only amplifies that problem because practically everybody now has an AI story: AI estimating, scheduling, document review, progress tracking, risk management, assistants and agents. Buyers aren’t starved for options. They’re drowning in them.

Which led me to ask Jeff the question I keep coming back to: with all this new technology, what actually separates the products that stick from the ones that disappear?

His first answer was refreshingly unsophisticated.

“It’s got to work… and the end users have the trust the data/responses. If they don't, it will quickly become shelfware and the pilot will fail.”

Fair enough.

But his second point was more interesting: it has to fit the way people actually work.

If people have to remember to use it, you may have already lost

Jeff told me about a conversation he once had with Mitch Cornelius, CTO at Fortis Construction. The idea that stuck with him was simple: if people have to consciously think about using a technology, there’s a pretty good chance they won’t.

“It has to just show up in their workflow,” Jeff says. That might mean a text telling someone there’s a submittal waiting for review, an alert that something changed or a task that needs action. The technology needs to come to the user rather than requiring the user to continually remember to go find the technology.

Because, as Jeff put it, “Construction is chaos.”

That immediately reminded me of something Steve Dell’Orto told me in another recent Buildernauts conversation: technology can’t make the job harder. It probably can’t even make the job the same. It needs to make the job easier.

Jeff had a story for that too.

Early in his time at Procore, he went to a construction trailer for a demonstration. Before Jeff and the Procore team came inside, the VDC manager had his team perform a simple task in the software they were already using: create an RFI.

Then he invited the Procore team inside and asked them to do exactly the same thing.

Jeff remembers it taking only a few clicks.

The customer essentially looked at them and said: What else?

Nothing else. That was the demo.

Jeff remembers the old workflow taking something like 17 clicks. The new one took a handful.

That story is probably more relevant to the AI boom than half the AI keynotes I’ve heard. Underneath all the talk about models, agents and automation is a remarkably basic question: Does this make the job easier?

If yesterday it took 17 clicks and today it takes three, that’s value. If AI turns those three clicks into one approval, even better. But if your incredible new AI application creates another login, another dashboard and another thing somebody in a construction trailer has to remember to check?

Good luck.

The best AI products may actually do less

There’s another misconception that seems to come with AI: because AI can theoretically do so many things, products should do more things.

Jeff thinks almost the opposite.

When he evaluates AI companies today, he looks for a clearly defined customer, a clearly defined problem, a quick path to value and what he calls a wedge.

“I don’t want to sell to everybody,” he told me.

He pointed to AI Clearing, a company working in progress tracking for horizontal construction. Jeff’s description of the value proposition was basically: Here’s what we were supposed to build. Here’s what we actually built. Do they match?

You don’t need a 30-slide deck to understand that. More importantly, the customer can potentially see value quickly.

The RICS data reinforces the importance of proving that value. Unclear ROI was cited by 28% of respondents as a barrier to AI adoption, and the survey found a 16-point gap between organizations planning increased AI investment and those reporting current active adoption. There is plenty of interest. Turning that interest into sustained usage is another matter.

And, increasingly, the alternative to buying a product isn’t simply doing nothing.

Jeff told me about Colin Stoner at Novo Construction, someone he considers one of the more sophisticated construction technologists he knows. Stoner had gone so far as to build his own version of Procore internally.

When Jeff was considering advising BuildCheck, an AI construction startup, he called Stoner because Novo was a customer and asked what he thought.

Stoner’s answer was simple:

“I can’t build BuildCheck.”

Jeff laughed when he told me the story.

“That’s all I needed to hear.”

I love that test.

In a world where the tools for building software are becoming more accessible by the month, the moat can’t simply be that you built some software. Increasingly, the question may be: Have you built something the customer can’t easily build for themselves?

The answer also has to be trusted

Jeff doesn’t actually think selling AI is fundamentally different from selling traditional software.

“You’ve got a defined problem and you’ve got a better mousetrap,” he said.

But there is one important difference. With traditional software, if the inputs are correct, users generally expect the system to produce a predictable result. AI introduces uncertainty.

Jeff framed it as the difference between deterministic and probabilistic systems. And that creates a new question for the buyer: Do I trust the answer?

That is a much bigger issue on a construction project than it is when I ask ChatGPT where I should eat dinner. If an AI system is reviewing a contract, evaluating progress, predicting schedule risk or providing information that influences a multimillion-dollar construction decision, “probably right” has a different meaning.

For now, Jeff thinks the answer is often some version of “trust, but verify.” The technology will get better, but companies may not get many chances to earn that trust. If someone in the field tries your product and the answer is wrong, they may decide the whole thing is vaporware.

And construction people have long memories.

Maybe adoption starts smaller than we think

One of my favorite examples from Jeff wasn’t about an AI startup at all.

He told me about Carl McFarland, president of Big-D’s Western Division, who challenged everyone on his team to find a way AI could have an impact on their work.

Not just the technology team. Everyone—from people at the front desk to people in the field.

And Jeff’s point wasn’t that everyone needed to go find some revolutionary enterprise AI platform. For someone, the first step might simply be using ChatGPT to help write an email.

“Find a way to make your life easier with AI. It’s a great way to baby-step into it.”

There’s something refreshingly practical about that.

We spend a lot of time talking about enterprise AI strategies, transformation roadmaps and autonomous agents. Those things matter. But widespread adoption may start with thousands of people finding one annoying part of their day and realizing there’s now an easier way to do it.

That’s also how most of us started using ChatGPT. Once a month became once a week, then once a day, and eventually it just became part of the workflow.

Which brings us right back to Jeff’s earlier point.

The best technology eventually stops feeling like technology you have to remember to use.

AI doesn’t eliminate the need for construction expertise

That led us into another conversation I’ve been having a lot lately: who builds the better construction technology company?

Someone who spent years in construction and experienced the problem firsthand? Or a brilliant technology team that works closely with construction customers to understand the problem?

Jeff’s answer was neither. The word he kept coming back to was empathy.

He told me about walking into one of his first construction trailers. The superintendent looked at him and immediately knew he hadn’t come from construction.

Jeff asked how he knew.

The guy looked down.

“You got zippers on your boots, bro.”

Welcome to construction, Jeff.

There’s a serious point behind the story. Jeff could run a sales process. What he couldn’t honestly tell a superintendent was, “When I was in your position, here’s how I would have handled this.”

That’s why he thinks some of the best ConTech companies surround their technologists and salespeople with people who have sat in those seats—sales engineers from the field, customer success people who understand the workflow, and industry advisors who can tell a product team that the feature they’re excited about doesn’t actually solve a problem anyone cares about.

AI doesn’t make that domain knowledge less important. I suspect it may make it more important.

As building software becomes easier, understanding which software is worth building becomes more valuable.

Nobody knows where this ends

Toward the end of our conversation, I asked Jeff where all this is going.

Specifically, I brought up an idea someone recently shared with me: what if the ultimate goal of adding AI to software is to make it so users barely need to use the software anymore?

Think about how much enterprise software consists of people opening an application, finding information, clicking through a workflow, moving something from one state to another and telling someone else that it happened. Agents could conceivably eliminate a lot of that. Five clicks become one. Eventually, perhaps the one click becomes an approval.

Jeff’s answer surprised me because he spends a ridiculous amount of time thinking about AI.

He said he doesn’t know. He doesn’t think anybody else really knows either.

“If anybody thinks they have the answer to you, they’re full of shit. The smartest people in the world don’t know the answer to that question.”

I appreciated that.

We can speculate about agents talking to agents. We can imagine workflows becoming increasingly automated. We can wonder what happens to huge systems of record such as Procore and Autodesk when the interface between humans and software fundamentally changes.

Jeff thinks that’s one of the biggest questions we’ll be looking back on five years from now. I think he’s right.

But after talking with him, I also wonder whether we’re making the future more complicated than it needs to be.

The technology is changing incredibly fast. The rules for winning may not be.

Solve a real problem. Know exactly who has that problem. Get them value quickly. Fit into the way they actually work. Make their job easier. Earn their trust. And perhaps most importantly, understand the human being who has to use the thing.

Twenty years ago, that might have meant turning 17 clicks into three. Today, AI might turn three into one.

Tomorrow? Who knows.

But if that one click doesn’t make someone’s life easier, I’m not sure any of the rest of it matters.

-WJ

Jeff Stieber brings 30 years of experience in technology, sales, and leadership, with a significant part of his career focused on construction technology. Throughout his career, he has held leadership roles and worked with companies including Procore, Briq, and Trunk Tools, helping drive growth and bring innovative technology to the construction industry. Today, through Stieber Advisory Group, Jeff works with founders and sales leaders at early-stage ConTech companies, providing go-to-market advising and recruiting to help them build their teams, scale revenue, and grow their businesses. He brings deep industry knowledge, practical sales experience, and a strong network across the ConTech ecosystem.